Retinal blood vessel segmentation algorithm based on improved U-Net

Yuan LIU; Baicheng LI; Chunbo WU

doi:10.3969/j.issn.1005-5630.202308280111

Journals >Optical Instruments >Volume 46 >Issue 5 >Page 9 > Article

Optical Instruments
Vol. 46, Issue 5, 9 (2024)

Retinal blood vessel segmentation algorithm based on improved U-Net

Yuan LIU, Baicheng LI^*, and Chunbo WU

Author Affiliations

School of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China

show less

DOI: 10.3969/j.issn.1005-5630.202308280111 Cite this Article

Yuan LIU, Baicheng LI, Chunbo WU. Retinal blood vessel segmentation algorithm based on improved U-Net[J]. Optical Instruments, 2024, 46(5): 9 Copy Citation Text

show less

Fig. 1. Structure of network model

Download full size | View in the Article

Fig. 2. Residual module

Download full size | View in the Article

Fig. 3. Detail enhancement model

Download full size | View in the Article

Fig. 4. Results of U-Net and our algorithm on DRIVE

Download full size | View in the Article

Fig. 5. Comparison of partial segmentation

Download full size | View in the Article

Fig. 6. Line chart for comparison of different algorithm indicators

Download full size | View in the Article

Fig. 7. Comparison of segmentation on DRIVE

Download full size | View in the Article

名称	参数配置
优化器	Adam
初始学习率	0.001
训练步数	200
批处理大小	4

Table 1. Experimental parameter configuration

名称	缩写	公式	含义
敏感性	Se	$ {\mathrm{Se}}=\dfrac{{\mathrm{TP}}}{{\mathrm{TP}}+{\mathrm{FN}}} $	血管被正确分割的指数
特异性	Sp	$ {\mathrm{Sp}}=\dfrac{{\mathrm{TN}}}{{\mathrm{TN}}+{\mathrm{FP}}} $	图像背景被正确分割的指数
准确性	Acc	$ {\mathrm{Acc}}=\dfrac{{\mathrm{TP}}+{\mathrm{TN}}}{{\mathrm{TP}}+{\mathrm{TN}}+{\mathrm{FP}}+{\mathrm{FN}}} $	图像整体被正确分割的指数
F1值	F1	${\mathrm{ F}}1=\dfrac{{\mathrm{2TP}}}{{\mathrm{2TP}}+{\mathrm{FP}}+{\mathrm{FN}}} $	衡量分割结果和标准结果之间相似性的指数

Table 2. Performance indicators

Algorithm	Acc	Se	Sp	F1
U-Net	0.9627	0.8159	0.9806	0.8137
U-Net+Res	0.9649	0.8168	0.9859	0.8192
U-Net+DEA	0.9659	0.8356	0.9831	0.8254
Our algorithm	0.9673	0.8373	0.9843	0.8293

Table 3. Results of different structural models based on U-Net

Method	Acc	Se	Sp	F1
AG-Net^[18]	0.9654	0.7638	0.9739	0.8132
CE-Net^[19]	0.9631	0.7746	0.9761	0.8156
DUNet^[20]	0.9579	0.7439	0.9872	0.8265
SCS-Net^[21]	0.9768	0.8157	0.9721	0.8241
CA-Net^[22]	0.9617	0.8042	0.9743	0.8104
ASU-Net^[23]	0.9678	0.8193	0.9789	0.8183
PVT-CASCADE^[24]	0.9736	0.8265	0.9773	0.8231
UNet-2022^[25]	0.9628	0.8073	0.9812	0.8117
Our method	0.9673	0.8373	0.9843	0.8293

Table 4. Results of different algorithms on DRIVE

Yuan LIU, Baicheng LI, Chunbo WU. Retinal blood vessel segmentation algorithm based on improved U-Net[J]. Optical Instruments, 2024, 46(5): 9

Download Citation

Tools

Save the article for my favorites

Paper Information

微信扫一扫：分享

微信扫一扫：分享